Now, the simple way to get a 32-bit input is left zero padding. However, in this case you are not adding any information and do not get any gain from a higher precision DAC. There is other one other thing that needs to be considered - Suppose that the digital pattern with all zeros corresponds to voltage 0.0 and that the digital pattern with all ones correspond to voltage 1.0V. In this case of zero-padded 24-bit word, the maximum voltage you are going to generate is (1/256)V. You may be playing with the dynamic range of your output!
The above is a problem if done the wrong way. And what I have explained is the wrong way
<Warning: Technical stuff>
Let me give an example of a 2-bit DAC converted to 4-bit DAC. The possible inputs with 2-bit DAC are "00", "01", "10" and "11" (binary equivalent of 0, 1, 2, 3). This would be converted to "0V", "0.25V", "0.5V" and "0.75V".
If we left pad them, the numbers are "0000", "0001", "0010" and "0011". With a 4-bit DAC the voltages become "0V", "0.0625V", "0.125V" and "0.1875V". This is the wrong way to do it.
However if we right pad, the numbers are "0000", "0100", "1000" and "1100" (equivalent decimal - 0, 4, 8, 12). A 4-bit DAC will give "0V", "0.25V", "0.5V" and "0.75V" as in the original case. This is the right way. But no advantage gained using 4-bit DAC.
</Technical stuff>
So then what do upsampling players such as CA 840C and Wadia players do? They upgrade the entire digital chain to produce higher precision digital samples at a higher rate. The computations are done at a higher precision. Also value are filled in between the time intervals using some intelligent algorithms. CA 840C input to the DAC is 32-bit precision at 384KHz while the Wadia does it at 24-bit precision at 1.4112MHz.
So, back to the original topic. I am a non-believer in modding until some the modder can scientifically explain me how he intends to improve the performance of a stock unit.
Case rested
Regards,
Prasad Redkar.